Reliability Optimization Design And Application Based On Surrogate Model

In the reliability optimization design of practical engineering,it is difficult to obtain either the optimization objective function of structural performance or the explicit expression of performance function.In addition,the calculation of the objective function optimization and reliability results in inaccurate calculations and low computational efficiency due to invocation of the actual object model.Based on this,this paper presents a reliability optimization design method based on multi-level target surrogate.In this paper,two-point addition strategy is used to improve the reliability calculation method,the objective function is approximated by a multi-level surrogate model technique,the adaptive dynamic penalty function method is used to deal with the reliability constraints,and the proposed method is applied to the reliability optimization engineering example of the clamping support of container back car.The main work of this paper is as follows:(1)Reliability calculation method based on two-point addition strategy.In order to select the appropriate surrogate model technology to approximate the real performance function,first of all,select several commonly used surrogate models and experimental design methods,and analyze the influence of the different combination modes on the reliability calculation results through the agent model response surface method.It is concluded that the Kriging model and the uniform design are the preferred combination method.Secondly,in the reliability optimization problem,the exact reliability calculation result is the key to ensure a feasible solution.Therefore,in order to obtain accurate reliability calculation results,an improved Kriging response surface method based on the two-point addition strategy is proposed.The checking point and the best sample point were added to the sample database of each iteration.The genetic algorithm was used to optimize the important parameters of the Kriging model,and the important sampling method was used to modify the reliability calculation results.(2)Reliability optimization design based on multi-level target surrogate.Multi-level target surrogate optimization method is proposed to find the optimal solution of the objective function.The Kriging model is used to fit the objective function in reliability optimization design.Based on the point selection criteria of Kriging prediction,the genetic algorithm is used for multi-point addition sequence optimization,and a local surrogate model is added to the optimal solution at each iteration to strengthen the local search and assist the Kriging model to global agent,then a multi-level surrogate model is constructed.The adaptive dynamic penalty function is used to deal with the reliability constraints.An improved Kriging response surface method based on two-point addition strategy is applied to carry out the reliability optimization analysis.(3)Reliability optimization design of clamping support of container back car.The reliability optimization design method based on the multi-level target surrogate is used to carry out the reliability optimization design of clamping support of container back car.Taking the minimum weight of clamping support as the optimization goal,the corresponding performance function reliability constraints are established according to the static strength failure mode,displacement failure mode and fatigue failure mode,and the reliability sensitivity analysis is performed using the Ansys Workbench 14.0 software to determine the significant basic random variables contained in each performance function with reliability constraints.The significant basic random variables included in the function are used to establish a mathematical model for reliability optimization.Under the condition that the reliability is 99.85%,the reliability optimization design method based on multi-level target surrogate is used to optimize the reliability of the structure.Comparing the three design options,the optimal design plan is optimized to achieve the lightweight design of the clamping support.The results show that this method can obtain satisfactory calculation accuracy and computational efficiency.